An Improved Hybrid Firefly Algorithm for Solving Optimization Problems

Abstract

The standard firefly algorithm is suffered from three major drawbacks. Firstly, imbalanced exploration and exploitation due to random initial solution generation. Secondly, the local convergence rate is low when the randomization factor is large. Thirdly, low quality local and global search capability at termination stage that result in failing to get the most optimal solution. To overcome all these drawbacks, a new approach is introduced which has been named GA-FA-PS algorithm in which genetic algorithm (GA) has been applied to generate the initial solution for balancing the exploration and exploitation at the initial stage. In the second stage, crossed over operator is embedded in firefly changing position to improve local search which ultimately enhances local convergence. To further improve the local and global convergence rate, pattern search (PS) is introduced which is used to obtain the most optimal solution or at least the solution better than the solution provided by the standard firefly algorithm. The performance of the proposed approach has been compared with standard FA and GA and the proposed method outperforms both of these approaches in terms solution quality.